WO2023120968A1

WO2023120968A1 - Method for predicting fractional flow reserve on basis of machine learning

Info

Publication number: WO2023120968A1
Application number: PCT/KR2022/017372
Authority: WO
Inventors: 하진용
Original assignee: 주식회사 레이와트
Priority date: 2021-12-21
Filing date: 2022-11-07
Publication date: 2023-06-29
Also published as: KR20230094295A; KR102656944B1

Abstract

A method for predicting fractional flow reserve on the basis of machine learning is disclosed. The disclosed method for predicting fractional flow reserve comprises the steps of: receiving a feature value for a target cardiac blood vessel, the feature value being extracted from an OCT image of a lumen of the target cardiac blood vessel and a class of the target cardiac blood vessel; and predicting a fractional flow reserve (FFR) value for the target cardiac blood vessel by using at least one pre-trained predictive model, the class of the target cardiac blood vessel, and the feature value, wherein the feature value includes at least one of a proximal lumea area (PLA) value, a minimal lumea area (MLA) value, a percent area stenosis value, a distal lumea area (DLA) value, a lesion length value, a rupture presence/absence value, an atheromatous plaque area value, and a lipid value.

Description

A method for predicting fractional blood flow reserve based on machine learning

The present invention relates to a method for predicting fractional blood flow reserve based on machine learning, and more particularly, to a learning method for predicting fractional blood flow reserve of a coronary artery patient and fractionation of a coronary artery patient using a prediction model obtained through such learning. It relates to a method for predicting blood flow reserve.

Fractional Flow Reserve (FFR) means the ratio of maximum blood flow in normal vessels distal to and proximal to coronary artery stenosis. Here, the proximal blood vessel means a blood vessel on the side of the coronary artery close to the heart, and the proximal blood vessel means a blood vessel on the side far from the heart.

A fractional blood flow reserve value of 0.75 at the coronary artery stenosis area means that coronary blood flow is reduced by up to 25% compared to normal coronary arteries.

As such, since the fractional blood flow reserve value decreases at the coronary artery stenosis site, whether or not coronary artery stenosis, that is, the cardiovascular system, is stenosis can be determined through the fractional blood flow reserve value, and a stent insertion procedure can be determined.

This fractional blood flow reserve value is the ratio of the pressure between the distal and proximal parts of the lesion measured using a pressure wire after maximal hyperemia was induced by intracoronary bolus injection or continuous intravenous injection of adenosine. can be calculated as a ratio.

Alternatively, as a non-invasive method, after extracting a lumen from a blood vessel cross-sectional image, a fractional blood flow reserve value may be measured through flow analysis based on a 3-dimensional cardiovascular model. Such an image-based method for measuring fractional blood flow reserve has a problem in that it is not easy to apply to an actual medical field due to a considerably long modeling and simulation time.

The present invention is to provide a faster and more accurate fractional blood flow reserve prediction method that can be used in the medical field.

In addition, the present invention is to provide a method for predicting fractional blood flow reserve capable of providing information necessary for a patient's stent operation.

According to one embodiment of the present invention for achieving the above object, the step of receiving feature values for the target cardiovascular class extracted from an OCT image of a class of the target cardiovascular and a lumen of the target cardiovascular; and predicting a fractional flow reserve (FFR) value for the target cardiovascular system using at least one pre-learned predictive model, the class of the target cardiovascular system, and the feature value, wherein the feature value is in the proximal blood vessel. A machine that includes at least one of area (PLA) value, minimum intravascular area (MLA) value, areal stenosis rate value, distal intravascular area (DLA) value, lesion length value, rupture presence/absence value, atherosclerotic plaque area value, and lipid value A learning-based method for predicting fractional blood flow reserve is provided.

In addition, according to another embodiment of the present invention for achieving the above object, generating a feature value for the target cardiovascular from an OCT image of the lumen of the target cardiovascular; and predicting a fractional blood flow reserve value for the target cardiovascular system using at least one pre-learned prediction model, the class of the target cardiovascular system, and the feature value, wherein the feature value includes a proximal intravascular area value, A method for predicting fractional blood flow reserve based on machine learning including at least one of minimum intravascular area value, area stenosis rate value, distal intravascular area value, lesion length value, rupture presence/absence value, atherosclerotic plate area value, and lipid value is provided.

According to one embodiment of the present invention, a more accurate fractional blood flow reserve prediction result can be provided using feature values obtained from OCT images.

In addition, according to one embodiment of the present invention, a more accurate prediction result can be provided by predicting the fractional blood flow reserve in consideration of the cardiovascular class for which the feature value has been obtained.

1 is a diagram for explaining a method for predicting fractional blood flow reserve based on machine learning according to an embodiment of the present invention.

2 is a diagram showing a predictive model according to an embodiment of the present invention.

3 is a diagram showing a predictive model according to another embodiment of the present invention.

Since the present invention can make various changes and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. Like reference numerals have been used for like elements throughout the description of each figure.

The present invention relates to a method for predicting the fractional blood flow reserve of a coronary artery patient using a pre-learned prediction model. According to the present invention, since the fractional blood flow reserve can be predicted more quickly and accurately through machine learning, can be applied

One embodiment of the present invention predicts the fractional blood flow reserve using feature values extracted from OCT images. Optical Coherence Tomography (OCT) is a technology that can image microstructures inside biological tissues by combining the light interference phenomenon and the principle of confocal microscopy, and shows higher resolution performance than ultrasound, CT, or MRI. . Therefore, according to an embodiment of the present invention for predicting the fractional blood flow reserve using the feature values extracted from the OCT image, a more accurate fractional flow reserve prediction result can be obtained.

The method for predicting fractional blood flow reserve according to an embodiment of the present invention may be performed in a computing device including a processor.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

Referring to FIG. 1 , the computing device according to an embodiment of the present invention receives a target cardiovascular class and characteristic values for the target cardiovascular system (S110). The feature values are feature values extracted from the OCT image of the lumen of the target cardiovascular system, such as a proximal lumen area (PLA) value, a minimum lumen area (MLA) value, and a percent area stenosis ) value, distal lumen area (DLA) value, lesion length value, rupture presence/absence value, plaque area value, and lipid rich value. .

The minimum intravascular area refers to the area of the vessel at the point where the width of the vessel lumen, that is, the width between the walls of the vessel, is minimum at the site of the narrowed lesion, and the proximal intravascular area refers to the area of the vessel from the point where the width of the lumen is minimum to the heart. represents the area of the blood vessel at the point showing the largest width among the widths of the lumen of . The distal intravascular area refers to the area of a blood vessel at a point where the width of the lumen is the largest in a direction away from the heart at a point where the width of the lumen is the smallest. The lesion length represents the length of a blood vessel from a point relative to the proximal intravascular area to a point relative to the distal intravascular area, and the area stenosis ratio represents the degree of stenosis of a blood vessel at the lesion site. In addition, the rupture presence or absence value indicates whether there is a rupture at the lesion site, and the atherosclerotic plate area value indicates the atherosclerotic plate area at the lesion site. Finally, the lipid value represents the amount of lipid at the lesion site.

The feature values may be generated in a separate computing device for extracting feature values or may be determined by a specialist analyzing an OCT image. Feature values can be generated from OCT images.

The computing device according to an embodiment of the present invention predicts the fractional blood flow reserve value for the target cardiovascular system using at least one pre-learned predictive model, the target cardiovascular class, and feature values (S120). Among several types of cardiovascular, the target cardiovascular from which the OCT image can be obtained may be different according to the patient's physical condition or condition. Therefore, the computing device uses the class of the target cardiovascular from which the OCT image was obtained to determine the fractional blood flow reserve value. predict

Depending on the embodiment, the computing device predicts one of values between 0 and 1 as the fractional blood flow reserve value, or outputs 0 when the predicted fractional blood flow reserve value is greater than or equal to a preset threshold value, for example, 0.6, and outputs 1 if the value is smaller than the threshold value. can

The fractional blood flow reserve force prediction model may be a prediction model based on various learning algorithms, or may be a prediction model based on a random forest algorithm or an artificial neural network.

In step S120, the computing device according to an embodiment of the present invention selects a prediction model corresponding to a target cardiovascular class from among prediction models for predicting fractional blood flow reserve values for cardiovascular classes of different classes, and selects the selected prediction model. Using this, the fractional blood flow reserve value can be predicted. Here, the cardiovascular class may be a left anterior descending artery (LAD), a left circumflex artery (LCX), or a right coronary artery (RCA).

Referring to FIG. 2, the predictive model is a model learned using training feature values for cardiovascular diseases of different classes, and the first prediction model 210 is a training feature value obtained from the left anterior descending limb and a fraction thereof. It is learned through the blood flow reserve value and predicts the fractional blood flow reserve value for the left anterior descending limb. The second prediction model 220 is learned through the training feature values obtained from the left circumference coronary artery and the fractional blood flow reserve value therefor, and predicts the fractional blood flow reserve value for the left circumference coronary artery. The third prediction model 230 is learned through training feature values obtained from the right coronary artery and the fractional blood flow reserve value therefor, and predicts the fractional blood flow reserve value for the right coronary artery.

At this time, the predictive models are learned through different groups of feature values for training. The first predictive model 210 is learned through a group of training feature values including area stenosis ratio, minimum intravascular area, lesion length, and rupture value obtained from the OCT image of the left anterior descending limb, and the second predictive model (220) is learned through a training feature group including area stenosis rate value, minimum intravascular area value, proximal intravascular area value, lesion length value, and atherosclerotic plaque area value obtained from the OCT image of the left circumferential coronary artery. . And the third predictive model 230 is a training feature value group including the minimum intravascular area value, area stenosis rate value, distal intravascular area value, atherosclerotic plaque area value, and lipid value obtained from the OCT image of the right coronary artery. learned through

When the target cardiovascular class is the left anterior descending artery, the computing device selects the first predictive model to predict the fractional blood flow reserve value, and when the target cardiovascular class is the left circumferential coronary artery, selects the second predictive model 220 Predict the value of fractional blood flow reserve. And, when the target cardiovascular class is the right coronary artery, the third predictive model 230 is selected to predict the fractional blood flow reserve value.

Further, the computing device predicts the fractional blood flow reserve force value by using the feature value group corresponding to the training feature value of each predictive model. That is, the computing device predicts the fractional blood flow reserve value using different feature value groups according to the target cardiovascular class. When the target cardiovascular class is the left anterior descending limb, the computing device sets a feature value group including an area stenosis rate value, a minimum intravascular area value, a lesion length value, and a rupture value obtained from an OCT image of the left anterior descending limb to a first predictive model ( 210), and if the class of the target cardiovascular system is left-circumferential coronary artery, a feature value group including area stenosis rate value, minimum intravascular area value, proximal intravascular area value, lesion length value, and atheromatous plate area value is determined. 2 Input to the predictive model (220). And, when the target cardiovascular class is the right coronary artery, a feature value group including the minimum intravascular area value, area stenosis rate value, distal intravascular area value, atheromatous plate area value, and lipid value obtained from the OCT image of the right coronary artery is input to the third prediction model 230.

Meanwhile, when the fractional blood flow reserve value is included in the set threshold range, the computing device according to an embodiment of the present invention may predict the fractional blood flow reserve value by adding the characteristic value to a feature value group, and the threshold range is the value of the target cardiovascular system. It can be determined by class.

In general, it is known that when the fractional flow reserve value for the lesion site is lower than 0.75, there is stenosis causing myocardial ischemia, and when the fractional flow reserve value is higher than 0.8, the stenosis does not cause myocardial ischemia. It is known that when the reserve power value is between 0.75 and 0.8, the presence or absence of myocardial ischemia due to stenosis is unknown. Therefore, in order to more accurately predict the fractional blood flow reserve value, when the fractional blood flow reserve value through the pre-prediction is included in a preset threshold range, the computing device generates additional feature values other than the existing feature values included in the feature value group. By inputting into the predictive model, the fractional blood flow reserve value can be predicted again. The computing device may additionally input one of the aforementioned feature values into the predictive model, and the additionally input feature value may be set in advance.

At this time, since there may be differences in accuracy for each predictive model, the computing device may additionally use a feature value according to a threshold range adjusted according to the target cardiovascular class. And, the width of the threshold range may be set in inverse proportion to the accuracy of the predictive model. For example, if the accuracy of the second prediction model 220 is the highest, the width of the threshold range is set to be the smallest, and if the accuracy of the third prediction model 230 is the lowest, the width of the threshold range is set to be the largest. can

As the number of generated feature values increases, the time and cost required to generate feature values increase. Therefore, in an embodiment of the present invention, when the uncertainty of the prediction result is high, the fractional blood flow reserve value is obtained by using additional feature values. predict

In step S120, the computing device may predict the fractional blood flow reserve by inputting the target cardiovascular class and feature values to the predictive model. Here, the feature value may include an area stenosis rate value, a minimum intravascular area value, and a distal intravascular area value.

Unlike the predictive model of FIG. 2 , the predictive model 310 of FIG. 3 is not separated by cardiovascular class, and predicts the fractional blood flow reserve value regardless of the cardiovascular class. Training data used for learning the predictive model of FIG. 3 includes not only training feature values and fractional blood flow reserve values therefor, but also cardiovascular classes from which training feature values were acquired.

Correspondingly, the computing device predicts the fractional blood flow reserve value by inputting not only the feature values of the target cardiovascular system but also the class of the target cardiovascular system into the prediction model 310 . For each target cardiovascular class, feature values input to the predictive model may all be the same.

The technical contents described above may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiments or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler. A hardware device may be configured to act as one or more software modules to perform the operations of the embodiments and vice versa.

As described above, the present invention has been described by specific details such as specific components and limited embodiments and drawings, but these are provided to help a more general understanding of the present invention, and the present invention is not limited to the above embodiments. , Those skilled in the art in the field to which the present invention belongs can make various modifications and variations from these descriptions. Therefore, the spirit of the present invention should not be limited to the described embodiments, and it will be said that not only the claims to be described later, but also all modifications equivalent or equivalent to these claims belong to the scope of the present invention. .

Claims

receiving feature values of the target cardiovascular class extracted from OCT images of a class of the target cardiovascular and a lumen of the target cardiovascular; and

Predicting a fractional blood flow reserve (FFR) value for the target cardiovascular system using at least one pre-learned prediction model, the class of the target cardiovascular system, and the feature value,

The feature values include a proximal intravascular area (PLA) value, a minimum intravascular area (MLA) value, an area stenosis rate value, a distal intravascular area (DLA) value, a lesion length value, a rupture presence value, an atherosclerotic plaque area value, and a lipid value. containing at least one of

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 1,

The step of predicting the fractional blood flow reserve is

selecting a prediction model corresponding to the class of the target cardiovascular class from among prediction models for predicting fractional blood flow reserve values for cardiovascular classes of different classes; and

predicting the fractional blood flow reserve value using the selected prediction model;

Machine learning-based fractional blood flow reserve prediction method comprising a.
According to claim 2,

The step of predicting the fractional blood flow reserve is

Predicting the fractional blood flow reserve value using different feature value groups according to the class of the target cardiovascular

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 3,

The step of predicting the fractional blood flow reserve is

When the target cardiovascular class is the left anterior descending artery (LAD), the fractional blood flow reserve is determined by using a feature value group including the area stenosis rate value, the minimum intravascular area value, the lesion length value, and the rupture presence/absence value. predicting the value

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 3,

The step of predicting the fractional blood flow reserve is

When the class of the target cardiovascular is left circumferential coronary artery (LCX), feature values including the area stenosis ratio value, the minimum intravascular area value, the proximal intravascular area value, the lesion length value, and the atheromatous plaque area value Predicting the fractional blood flow reserve value using a group

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 3,

The step of predicting the fractional blood flow reserve is

When the target cardiovascular class is right coronary artery (RCA), a feature value group including the minimum intravascular area value, the area stenosis rate value, the distal intravascular area value, the atheromatous plate area value, and the lipid value Using, predicting the fractional blood flow reserve value

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 3,

The step of predicting the fractional blood flow reserve is

When the fractional blood flow reserve value is within a preset threshold range, predicting the fractional blood flow reserve value by adding the characteristic value to the feature value group;

The threshold range is determined according to the class of the target cardiovascular

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 1,

The step of predicting the fractional blood flow reserve is

Predicting the fractional blood flow reserve value by inputting the target cardiovascular class, the area stenosis rate value, the minimum intravascular area value, and the distal intravascular area value into the prediction model

A method for predicting fractional blood flow reserve based on machine learning.
generating feature values for the target cardiovascular system from an OCT image of a lumen of the target cardiovascular system; and

Predicting a fractional blood flow reserve value for the target cardiovascular system using at least one pre-learned predictive model, the class of the target cardiovascular system, and the feature value;

The feature value includes at least one of a proximal intravascular area value, a minimum intravascular area value, an area stenosis rate value, a distal intravascular area value, a lesion length value, a rupture presence value, an atherosclerotic plaque area value, and a lipid value.

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 9,

The step of predicting the fractional blood flow reserve is

selecting a prediction model corresponding to the class of the target cardiovascular class from among prediction models for predicting fractional blood flow reserve values for cardiovascular classes of different classes; and

predicting the fractional blood flow reserve value using the selected prediction model;

Machine learning-based fractional blood flow reserve prediction method comprising a.
According to claim 10,

The step of predicting the fractional blood flow reserve is

Predicting the fractional blood flow reserve value using different feature value groups according to the class of the target cardiovascular

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 11,

The step of predicting the fractional blood flow reserve is

When the class of the target cardiovascular system is the left anterior descending artery, predicting the fractional blood flow reserve value using a feature value group including the area stenosis rate value, the minimum intravascular area value, the lesion length value, and the rupture presence/absence value

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 11,

The step of predicting the fractional blood flow reserve is

When the class of the target cardiovascular is a left-circumferential coronary artery, a feature value group including the area stenosis rate value, the minimum intravascular area value, the proximal intravascular area value, the lesion length value, and the atheromatous plate area value is used. So, to predict the fractional blood flow reserve

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 11,

The step of predicting the fractional blood flow reserve is

When the target cardiovascular class is the right coronary artery, using a feature value group including the minimum intravascular area value, the area stenosis rate value, the distal intravascular area value, the atherosclerotic plate area value, and the lipid value, Predicting the fractional blood flow reserve

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 11,

The step of predicting the fractional blood flow reserve is

When the fractional blood flow reserve value is within a preset threshold range, predicting the fractional blood flow reserve value by adding the characteristic value to the feature value group;

The threshold range is determined according to the class of the target cardiovascular

A method for predicting fractional blood flow reserve based on machine learning.
According to claim 9,

The step of predicting the fractional blood flow reserve is

Predicting the fractional blood flow reserve value by inputting the cardiovascular class, the area stenosis rate value, the minimum intravascular area value, and the distal intravascular area value into the prediction model

A method for predicting fractional blood flow reserve based on machine learning.